Step I: peak detection
The open-source software MZmine (version 2.3) was applied to
automatically deconvolute the full scan data and to generate peak
lists. The acquired data was constrained to a mass to charge ratio
(m/z) range of 100–423 Da based on observations by Helbling et al.(2010) who concluded that it is not likely for a parent compound
with m/z of a to have a TP with m/z > (50 + a) Da [10]. During the peak
picking process, each mass with intensity above a defined noise
level (threshold: 10) was used to build a chromatogram within an
m/z window of 20 mDa. The chromatograms were smoothed by the
Savitzky–Golay method (window size: 5 data points) before they
were deconvoluted to generate peak lists containing m/z, retention
time (RT) and peak intensity of each feature.
The enviMass workflow (version 1.0) [15] was applied to process
the peak lists. For the two batches of experiments with river
Roter Main sediment – for which a blank-control experiment
was available – the peak lists of the incubations that contained
pharmaceuticals and the corresponding blank-control at the same
sampling time were simultaneously imported into enviMass and
the blank subtraction routine was applied (tolerance windows:
m/z: ±5 mDa; RT: ±0.05 min). Peaks with an at least four times
higher intensity compared to the blank-control were retained to
keep the initially spiked pharmaceuticals for further processing.
For the experiment with river Gründlach sediment, the blank subtraction
step was omitted due to the lack of a blank experiment.
Subsequently, the parent pharmaceuticals and internal standards
were filtered using m/z ratios, isotopic patterns, and RT. Finally,
peaks with mass differences equal to 13C and 37Cl isotopologue
ions and RT within a window of ±0.05 min were assigned to distinguish
the components with monoisotopic mass. Only those with an
observed 13C isotope peak were consequently retained in the list
of “unknowns” for further processing